Cardiovascular risk assessment: baseline snapshots or accumulated burden?

This editorial refers to ‘Effects of individual and integrated cumulative burden of blood pressure, glucose, LDL-C, and CRP on cardiovascular risk’, by Z. Chen et al., pp. 127–135. Cardiovascular morbidity is a growing public health concern. Guidelines suggest preventive interventions for individuals at high risk for cardiovascular events,1 and over 30% of the population aged 40–75 years may be eligible for statin therapy according to current European guidelines.2 A 2016 meta-analysis found 363 papers that described the development of different cardiovascular risk prediction models.3 This great abundance reflects a pressing need for accurate, individually tailored risk estimation tools to guide therapy. While cardiovascular damage evolves over prolonged periods, current prediction models use only a ‘snapshot’ of baseline data as a proxy for actual long-term exposure to risk factors. Furthermore, within-person variability and measurement errors may bias risk prediction if based on a single measurement. This suggests a need to consider both the magnitude and the cumulative duration of exposure to risk factors. The authors of ‘Effects of individual and integrated cumulative burden of blood pressure, glucose, LDL-C, and CRP on cardiovascular risk’ proposed to explore the effect of the cumulative burden of cardiovascular risk factors on cardiovascular events. In this issue of the European Journal of Preventive Cardiology, Chen et al. report a secondary analysis of a cohort of 35 000 employees of the Kailuan Company in Tangshan city in the north of China; comprising 34% of the participants of the larger Kailuan Study, who had completed all four biannual surveys from 2006 to 2013, had available data to assess risk factors, and had not experienced a cardiovascular event before the 2012–2013 survey (the index date). They used self-reported as well as clinical and laboratory data from these surveys to calculate a metric of cumulative cardiovascular risk (for each of several established risk factors separately as well as a pooled metric for all of them). They then used Cox regression models to assess whether the degree of cumulative burden was associated with cardiovascular events and all-cause mortality during the 5 years following the 2012–2013 survey. The authors utilized the region’s routinely collected data to ascertain outcomes (847 ischaemic strokes, 221 myocardial infarctions, and 1185 deaths). They demonstrated that those at the upper quartile of cumulative burden of cardiovascular risk factors were 2.45 (2.03–2.94) times more likely to experience any cardiovascular event compared to those at the lowest quartile [hazard ratio 3.65 (2.68–4.96) for an ischaemic stroke, 4.51 (2.19–9.27) for myocardial infarction, and 1.73 (1.36–2.21) for all-cause mortality]. Chen et al. also showed a dose–response association between cumulative risk factor burden and cardiovascular outcomes, supporting a causal association. These results remained robust through a range of sensitivity and stratified analyses. In a final analysis, Chen et al. reported that the c-statistic of the models predicting an outcome was higher (c = 0.67) if including the cumulative burden of risk factors compared to using only ‘snapshot’ baseline information (c = 0.58), implying improved prediction. The authors concluded that cardiovascular risk factors’ cumulative burden was a reliable predictor for cardiovascular outcomes, suggesting the necessity to extend the observation duration when measuring risk factors. Chen et al. were able to follow a large community-based cohort, using routinely collected information from local hospitals and a mortality register to capture outcomes with minimal loss to follow-up. However, several limitations of this study should be acknowledged. The original Kailuan study included over a hundred thousand participants, only half of whom were eligible to participate in the current analysis (i.e. those who completed all four surveys). Over 11 000 eligible participants were further excluded due to missing data. These people were more likely to be less physically active, drink alcohol, have diabetes, hypertension, and hypercholesterolaemia. Furthermore, the Kailuan cohort is a workers’ cohort, suggesting selection bias towards healthy and fit individuals (i.e. healthy workers effect), raising questions about the generalizability of the results. Measuring the cumulative burden of cardiovascular risk factors is not straightforward; there is no consensus regarding the methodology to do so. Chen et al. chose to calculate the cumulative average levels of several available markers (systolic blood pressure, fasting blood glucose, low-density lipoprotein cholesterol levels, and C-reactive protein) and categorize these into quartiles. The burden of each risk factor, i.e. the risk quartile, was pooled into an integrated score (with potential values between 4 and 16). This intuitive approach assigns a similar weight to all risk factors, most likely an inaccurate assumption. They also could not obtain data on the cumulative exposure to various established risk factors, such as physical activity, smoking, and body mass index; and had to rely on the ‘snapshot’ baseline levels of these factors in their models. Previous studies employed different methodologies to assess the prognostic usefulness of repeated measurement of cardiovascular risk factors, with varying results. A large 2017 study explored the added value of incorporating repeated blood pressure and cholesterol measurements to predict cardiovascular disease. Using time-updated post-baseline measurements, the authors demonstrated only a slight improvement in the ability to predict cardiovascular outcomes.4 Conversely, an analysis of Finnish data with a follow-up of 50 years suggested that using averages of risk factor measurements, rather than only the most recent information resulted in better predictive ability for cardiovascular deaths.5 A study which modelled various approaches to incorporate repeated blood pressure measurements to predict cardiovascular risk in the ARIC cohort concluded that using the cumulative mean of systolic blood pressure resulted in a modest improvement in risk prediction, and found little further gain when employing more complex modelling approaches.6 The study by Chen et al. addresses an interesting clinical question, i.e. should one take into account the cumulative nature of risk factor exposure and how should this be done. The study highlights the need to elucidate ways to use accumulated medical information to inform clinical decisions. Recent advances in handheld and mobile technology essentially place a continuous monitoring device in the entire population's pocket. Efforts to include continuous measures of risk factors using mobile apps are increasingly reported.7,8 Worldwide, many health systems record all information electronically, making a compelling case for using routinely collected electronic health information to furthers assess the utility of repeated and cumulative measurements of risk factors. The steep rise in available information will increase the demand for more precise, individually tailored, prediction models, but more research is needed to explore and refine the modelling strategies. Conflict of interest: none declared. The opinions expressed in this article are not necessarily those of the Editors of the European Heart Journal or of the European Society of Cardiology.